Remove all comparison-related code used to evaluate XML vs library-based
Office document extraction. The XML approach has proven superior, so the
comparison functionality is no longer needed.
Changes:
- Remove extraction_comparator.rs (entire comparison engine)
- Remove test_extraction_comparison.rs binary
- Remove comparison mode logic from enhanced.rs
- Simplify fallback_strategy.rs to use XML extraction only
- Update OCR service to use XML extraction as primary method
- Clean up database migration to remove comparison-specific settings
- Remove test_extraction binary from Cargo.toml
- Update integration tests to work with simplified extraction
The Office document extraction now flows directly to XML-based
extraction
without any comparison checks, maintaining the superior extraction
quality
while removing unnecessary complexity.
✅ Core Optimizations Implemented
1. 📊 New Database Schema: Added webdav_directories table to track
directory ETags, file counts, and metadata
2. 🔍 Smart Directory Checking: Before deep scans, check directory
ETags with lightweight Depth: 0 PROPFIND requests
3. ΓÜí Skip Unchanged Directories: If directory ETag matches, skip the
entire deep scan
4. 🗂️ N-Depth Subdirectory Tracking: Recursively track all
subdirectories found during scans
5. 🎯 Individual Subdirectory Checks: When parent unchanged, check
each known subdirectory individually
🚀 Performance Benefits
Before: Every sync = Full Depth: infinity scan of entire directory
treeAfter:
- First sync: Full scan + directory tracking setup
- Subsequent syncs: Quick ETag checks → skip unchanged directories
entirely
- Changed directories: Only scan the specific changed subdirectories
📁 How It Works
1. Initial Request: PROPFIND Depth: 0 on /Documents → get directory
ETag
2. Database Check: Compare with stored ETag for /Documents
3. If Unchanged: Check each known subdirectory (/Documents/2024,
/Documents/Archive) individually
4. If Changed: Full recursive scan + update all directory tracking
data
perf3ct
aaldebs99